Add hybrid AI-assisted model recommendations

[codoom1]

What changed

• keeps the deterministic statistical scoring engine as the first-stage shortlist generator
• adds an optional, checked-by-default AI review for authenticated users
• lets OpenAI choose only among verified shortlisted models and explain fit, assumptions, and alternative tradeoffs
• uses strict Responses API structured output plus server-side allowlist validation
• consumes one durable per-user AI quota unit per review
• falls back to the complete rules-based recommendation when AI, quota storage, or provider access is unavailable
• adds clear AI-assisted and fallback states to the results page

Why

The application previously used OpenAI for the chatbot and questionnaire tools, but not for the recommendation itself. This adds a bounded hybrid layer without allowing AI to invent models or making core recommendations dependent on the provider.

User impact

Signed-in users can request a richer AI review directly from the model-selection form. Guests and users who disable the option continue to receive the deterministic recommendation. A failed AI request never blocks the result page.

Validation

• python -m pytest -q tests/ — 128 passed
• focused flake8 fatal-error checks — passed
• git diff --check — passed
• Chrome mobile emulation at 390px — no horizontal overflow (scrollWidth == innerWidth == 390)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
statistical-model-suggester	Ready	Preview, Comment	Jul 25, 2026 12:00am

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.